Demand side adjustable resource response optimization method based on model predictive control theory

Abstract

The invention relates to the technical field of power grid supply and demand adjustment, in particular to a demand side adjustable resource response optimization method based on a model predictive control theory, which comprises the following steps: analyzing an off-grid condition of a load end to obtain a preposed interval duration of a non-rigid load; obtaining an off-grid span aggregation degree of each non-rigid load; obtaining a transferable load attribution degree by using a covariant relationship between the preposed interval duration and the consumed electric energy of the compensation power supply station; analyzing the complexity of the power grid topology network to obtain the power supply priority of each non-rigid load; a decision variable is obtained through model prediction; analyzing the difference between the off-network duration and the preposed interval duration by using the decision variable to obtain a decision update variable; and the power grid determines grid connection and grid disconnection of each non-rigid load in each time period in the prediction time domain by using the decision update variable. According to the method, the power supply cost is combined during resource adjustment, the off-grid duration of the transferable load is considered, and the safety of the power grid is improved by setting the priority of model prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid supply and demand regulation, and in particular to a demand side adjustable resource response optimization method based on model predictive control theory. BACKGROUND

[0002] In the face of the volatility challenge brought by the high proportion of renewable energy connected to the grid, the power system urgently needs to tap the response potential of demand side adjustable resources to maintain real-time balance. The mixed integer linear programming, according to this demand, formulates a precise response strategy by constructing a mathematical optimization framework, taking the predicted information of load demand, energy storage discharge power, etc. as input, defining the device start-stop state as an integer variable, and under the premise of strictly meeting the power balance and device operation constraints, rolling optimization is carried out to minimize the total operation cost, so as to solve the optimal coordinated control sequence of each load and energy storage device in the future prediction time domain, and continuously correct the strategy based on the feedback of actual response, and finally realize the economic and reliable guidance of demand side resources to actively support the operation of the power grid. The optimal coordinated control sequence represents the on-off state of the load or energy storage device at each time period in the prediction time domain. The device start-stop state is usually represented by binary decision variables. When the device is turned on at a certain time period, the decision variable is represented by 1, and vice versa.

[0003] Since the demand side load is mainly divided into rigid load and non-rigid load, the rigid load is the load that must be maintained during resource scheduling and regulation, and the non-rigid load is mainly divided into interruptible load and transferable load. The interruptible load can be disconnected from the grid at any time under the premise that the output power of the energy storage end is insufficient, and the transferable load can be transferred into the grid to maintain the stability of the transferable load. However, the transferable load needs to be reconnected to the grid within a certain time after being disconnected from the grid. If the transferable load is disconnected from the grid for a long time, it will cause safety risks or production losses to the transferable load, and significantly reduce the comfort of the user of the transferable load. SUMMARY

[0004] The present application provides a demand side adjustable resource response optimization method based on model predictive control theory to solve the existing problems.

[0005] The demand side adjustable resource response optimization method based on model predictive control theory of the present application adopts the following technical scheme: One embodiment of the present application provides a demand side adjustable resource response optimization method based on model predictive control theory, which comprises the following steps: Obtain the historical power sequence of each load end and the state of charge of each energy storage end at the current time, and construct the power grid topology network of the load end and the energy storage end according to the layout of the load end and the energy storage end in the power grid; The off-grid situation of each load end is analyzed by using a historical power sequence, the load end is divided to obtain a plurality of non-rigid loads, and each working power period and its preceding interval duration of each non-rigid load are obtained; the off-grid span aggregation degree of each non-rigid load is obtained according to the consistency of all the preceding interval durations; the transferable load attribution degree of each non-rigid load is obtained by using the covariant relationship between each preceding interval duration and the total amount of consumed electric energy in compensating power supply under the corresponding working power period; The complexity of the power grid topology network is analyzed to obtain the power supply cost of each non-rigid load for all energy storage ends; the power supply priority of each non-rigid load is obtained by using the transferable load attribution degree and the power supply cost; the decision variable of each non-rigid load in each period within the prediction time domain is obtained by using the power supply priority and the state of charge of all energy storage ends at the current moment for model prediction; At each period within the prediction time domain, the difference between the off-grid duration of each non-rigid load and the preceding interval duration thereof is analyzed by using the decision variable, and the decision update variable of each non-rigid load in each period within the prediction time domain is obtained by combining the real-time state of charge of the energy storage end; the power grid determines the grid connection and off-grid of each non-rigid load at each period within the prediction time domain by using the decision update variable.

[0006] Preferably, the step of analyzing the off-grid situation of each load end by using a historical power sequence, dividing the load end to obtain a plurality of non-rigid loads, and obtaining each working power period and its preceding interval duration of each non-rigid load comprises: The average value of the power at all moments in the historical power sequence of the i th load end is denoted as the historical power average value of the i th load end; A preset off-grid coefficient is multiplied by the historical power average value of the i th load end to obtain the low-power threshold value of the i th load end; All moments with power less than the low-power threshold value in the historical power sequence of the i th load end are denoted as the low-power moments of the i th load end; According to the proportion of the number of low-power moments of each load end in the historical power sequence of the load end, all load ends are divided to obtain non-rigid loads and rigid loads, and each working power period and its preceding interval duration of each non-rigid load are obtained.

[0007] Preferably, the step of dividing all load ends to obtain non-rigid loads and rigid loads according to the proportion of the number of low-power moments of each load end in the historical power sequence of the load end, and obtaining each working power period and its preceding interval duration of each non-rigid load comprises: The calculation method of the non-rigid load performance degree of the i th load end is as follows: ​ wherein, is a time length of the historical power sequence, is a number of low-power time instants in the historical power sequence of the i-th load end; a preset rigid load screening threshold; if the non-rigid load performance degree of the i-th load end is greater than or equal to the rigid load screening threshold, the i-th load end is recorded as a non-rigid load, each interval composed of non-low-power time instants in the historical power sequence of the non-rigid load is recorded as a work power time period of the non-rigid load, and a time length between the last time instant of a preceding work power time period of each work power time period and the first time instant of the work power time period in the historical power sequence of the non-rigid load is recorded as a preceding interval length of each work power time period of the non-rigid load.

[0008] Preferably, the obtaining of the off-grid span concentration of each non-rigid load according to the consistency of all the preceding interval lengths comprises: for the j-th non-rigid load, the preceding interval lengths of all the work power time periods of the j-th non-rigid load are counted to obtain a kurtosis of all the preceding interval lengths, denoted as . a range of the preceding interval lengths of all the work power time periods of the j-th non-rigid load is recorded as . the calculation manner of the off-grid span concentration of the j-th non-rigid load is: .

[0009] Preferably, the obtaining of the transferable load attribution degree of each non-rigid load by using the covariant relationship between each preceding interval length and the total amount of consumed electric energy in compensatory power supply in the corresponding work power time period comprises: a compensatory analysis space of the j-th non-rigid load is constructed, wherein an X axis is a time length of each work power time period of the j-th non-rigid load, a Y axis is a preceding interval length of each work power time period of the j-th non-rigid load, and a Z axis is a total output electric energy of each work power time period of the i-th non-rigid load, the total output electric energy being expressed as an integral of time over power in each work power time period of the j-th non-rigid load; the time length, the preceding interval length and the total output electric energy of all the work power time periods of the j-th non-rigid load are mapped into the compensatory analysis space to obtain a plurality of sample points in the compensatory analysis space of the j-th non-rigid load; a least square method is used to linearly fit all the sample points in the compensatory analysis space to obtain a compensatory straight line of the i-th non-rigid load; analyzing a fitting correlation degree of each sample point in a compensation analysis space of the jth non-rigid load with the compensation straight line to obtain a compensation performance degree of the jth non-rigid load; obtaining a transferable load attribution degree of the jth non-rigid load according to the compensation performance degree of the jth non-rigid load and the off-grid span aggregation degree of the jth non-rigid load, wherein the transferable load attribution degree is in a positive proportional relationship with the compensation performance degree and the off-grid span aggregation degree.

[0010] Preferably, the step of analyzing the fitting correlation degree of each sample point in the compensation analysis space of the jth non-rigid load with the compensation straight line to obtain the compensation performance degree of the jth non-rigid load comprises: The calculation manner of the step of analyzing the fitting correlation degree of each sample point in the compensation analysis space of the jth non-rigid load with the compensation straight line to obtain the compensation performance degree of the jth non-rigid load is: the compensation performance degree of the jth non-rigid load The specific manner is: wherein, the number of work power time periods of the jth non-rigid load, the Euclidean distance of a sample point corresponding to the kth work power time period in the compensation analysis space of the jth non-rigid load from the compensation straight line of the jth non-rigid load; an exponential function with a natural constant as a base, used for inversely proportional normalization of the accumulated sum of the Euclidean distances.

[0011] Preferably, the step of obtaining the power supply priority of each non-rigid load by using the transferable load attribution degree and the power supply cost comprises: obtaining a shortest path from the mth energy storage end to the jth non-rigid load in the power grid topology network; obtaining a cost contribution coefficient of each energy storage end to the jth non-rigid load according to the shortest path; taking an average value of the cost contribution coefficients of all energy storage ends to the jth non-rigid load as a power supply cost of each non-rigid load by all energy storage ends ; obtaining the power supply priority of each non-rigid load by using the transferable load attribution degree and the power supply cost, wherein the power supply priority is in a positive proportional relationship with the transferable load attribution degree and in a positive proportional relationship with the power supply cost.

[0012] Preferably, the specific obtaining step of the cost contribution coefficient comprises: The cost contribution coefficient of the mth energy storage end to the jth non-rigid load The calculation method is as follows: Wherein, is the number of nodes on the shortest path from the mth energy storage end to the jth non-rigid load; is the state of charge of the mth energy storage end at the current moment; is the cumulative sum of the degrees between all adjacent nodes on the shortest path from the mth energy storage end to the jth non-rigid load; is the prior art.

[0013] Preferably, the difference between the off-grid time length of each non-rigid load and the pre-interval time length thereof at each time period within the prediction time domain is analyzed by using the decision variable, and the real-time state of charge of the energy storage end is combined to obtain the decision update variable of each non-rigid load at each time period within the prediction time domain, which includes: The difference between the off-grid time length of each non-rigid load and the pre-interval time length thereof at each time period within the prediction time domain is analyzed by using the decision variable, and the supply priority of each non-rigid load at each time period within the prediction time domain is adjusted to obtain the adjusted priority of each non-rigid load at each time period within the prediction time domain. The method for obtaining the decision variable of each non-rigid load at each time period within the prediction time domain is used, and the real-time state of charge of all energy storage ends at each time period within the prediction time domain and the adjusted priority are taken as inputs to obtain the decision update variable of each non-rigid load at each time period within the prediction time domain.

[0014] Preferably, the specific obtaining step of the adjusted priority includes: The average of the pre-interval time lengths of all work power periods of the jth non-rigid load is denoted as ; Any time period within the prediction time domain except the first time period is denoted as the target time period q; The calculation method of the adjusted priority of the jth non-rigid load at the target time period q within the prediction time domain is as follows: Wherein, is the number of time periods with a decision variable of 1 before the qth time period within the prediction time domain, is the transferable load attribution degree of the jth non-rigid load; is the adjusted priority of the jth non-rigid load at the time period before the target time period within the prediction time domain, is the preset time length of the time period.

[0015] The beneficial effects of the technical solutions of the present application are: the present application obtains the historical power sequence of the load end and the state of charge of the energy storage end, constructs the power grid topology network of the load end and the energy storage end, analyzes the off-grid situation of each load end by using the historical power sequence to obtain a plurality of non-rigid loads, and obtains the work power period and its pre-interval duration; by analyzing the number of times each load end is off-grid in the historical power sequence, the load end is divided, so as to preferentially meet the rigid load when adjusting the resource; according to the consistency of all the pre-interval durations, the off-grid span aggregation degree of each non-rigid load is obtained; the covariant relationship between each pre-interval duration and the total amount of consumed electric energy when compensating for power supply in the corresponding work power period is used to obtain the transferable load attribution degree of each non-rigid load; by analyzing the maximum interval duration of the transferable load in the non-rigid load, and the electric energy compensation required according to the off-grid duration after the transferable load is reconnected to the grid, the transferable load attribution degree of the non-rigid load is obtained; the complexity of the power grid topology network is analyzed to obtain the power supply cost of all energy storage ends to each non-rigid load; the transferable load attribution degree and the power supply cost are used to obtain the power supply priority of each non-rigid load; by considering the power supply cost of the energy storage end to the non-rigid load in the power grid, and combining the transferable load attribution degree to preferentially supply power to the transferable load, the power supply priority of the non-rigid load is obtained; the power supply priority and the state of charge of all energy storage ends at the current time are used for model prediction to obtain the decision variable of each non-rigid load at each period in the prediction time domain; at each period in the prediction time domain, the difference between the off-grid duration of each non-rigid load and the pre-interval duration thereof is analyzed by using the decision variable, and the real-time state of charge of the energy storage end is combined to obtain the decision update variable of each non-rigid load at each period in the prediction time domain; after obtaining the decision variable of each non-rigid load by using model prediction, the length of time that the transferable load is off-grid from the power grid is considered, and the power supply priority is adjusted to avoid the transferable load from off-grid for a long time to cause the stability of the power grid to decrease; the power grid uses the decision update variable to determine the grid connection and off-grid of each non-rigid load at each period in the prediction time domain, and the present application sets the priority of model prediction by combining the power supply cost and considering the off-grid duration of the transferable load when adjusting the resource, thereby improving the safety of the power grid. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creating labor.

[0017] Figure 1A flow chart of steps of the demand side adjustable resource response optimization method based on model predictive control theory of the present application. DETAILED DESCRIPTION

[0018] In order to further clarify the technical means and effects taken by the present application to achieve the predetermined inventive purpose, the specific embodiments, structures, features and effects of the demand side adjustable resource response optimization method based on model predictive control theory according to the present application are described in detail below in combination with the drawings and preferred embodiments. Different "one embodiment" or "another embodiment" in the following description do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0020] The specific scheme of the demand side adjustable resource response optimization method based on model predictive control theory provided by the present application is specifically described below in combination with the drawings.

[0021] Please refer to Figure 1 which shows a flow chart of steps of the demand side adjustable resource response optimization method based on model predictive control theory provided by one embodiment of the present application, which includes the following steps: Step S001, obtain the historical power sequence of each load end and the state of charge of each energy storage end at the current time, and construct the power grid topology network of the load end and the energy storage end according to the layout of the load end and the energy storage end in the power grid.

[0022] It should be noted that the response and scheduling of adjustable resources in the power grid is to maintain the overall stability of the power grid by interrupting interruptible loads and intervening and regulating the energy storage end to provide as much power as possible for transferable loads under the premise of providing power support, under the premise of meeting the load of rigid load; by converting the actual power grid scheduling into a mathematical programming prediction model based on mixed integer linear programming, each demand end and energy storage end in the prediction time domain is predicted, so as to obtain the optimal collaborative control sequence of each power supply and consumption unit, and by controlling the opening and closing of the power supply and consumption unit, the stability of the output and consumption of power in the power grid is maintained, and a feedback correction mechanism is introduced in each period of the prediction time domain, and the optimal collaborative control sequence is adjusted in real time through the feedback correction mechanism to respond to the real-time change of power in the power grid.

[0023] Based on the above, the embodiment first needs to obtain the historical power sequence of each load end and the state of charge of each energy storage end at the current time, for predicting the input and output power in the prediction time domain, and constructing the power grid topology network according to the layout of the load end and the energy storage end in the power grid.

[0024] Specifically, the historical power sequence of each load end and the state of charge of each energy storage end at the current time are obtained, and the specific steps of constructing the power grid topology network of the load end and the energy storage end according to the layout of the load end and the energy storage end in the power grid are as follows: The power at each time is obtained through the power detection module of each load end and recorded in the log of each load end, and the power at each time in the historical time range is read from the log of each load end to constitute the historical power sequence of each load end; The ratio of the remaining power of each energy storage end to the rated total capacity at the current time is recorded as the state of charge of each energy storage end at the current time; According to the layout in the power grid, each load end and energy storage end in the power grid is taken as a node, and the length of the line between any two nodes is taken as the degree between the two nodes, to construct the power grid topology network of the load end and the energy storage end; if there is no line connection between two nodes, the corresponding two nodes in the power grid topology network are not connected; It should be noted that the time in the embodiment is taken as an example of one second, and the historical time range is taken as an example of 30 days, and other embodiments can use other values, which are not limited in the embodiment.

[0025] Step S002, analyze the off-grid situation of each load end using the historical power sequence, divide the load end to obtain a plurality of non-rigid loads, and obtain each work power period and its pre-interval length of each non-rigid load; according to the consistency of all the pre-interval lengths, obtain the off-grid span aggregation degree of each non-rigid load; obtain the transferable load attribution degree of each non-rigid load using the covariant relationship between each pre-interval length and the total amount of consumed electric energy in the corresponding work power period.

[0026] It should be noted that the load end mainly includes rigid load, transferable load and interruptible load, wherein the rigid load is the most basic load such as industrial core load and maintenance load, and interrupting the rigid load will bring great negative feedback to the power grid, so when the adjustable resource is scheduled according to the demand end, the power supply of the rigid load needs to be met first; when the power in the power grid can only support the operation of the rigid load, the transferable load and the interruptible load need to be off-grid in the power grid to ensure the carrying of the rigid load, and when there is still surplus power after supplying the rigid load, the transferable load and the interruptible load can be started according to the surplus power. Therefore, when the adjustable resource is scheduled according to the demand end, it is necessary to distinguish the rigid load from the non-rigid load in the power grid.

[0027] It should be further noted that the transferable load has fixed running power in the power grid, flexible running time period, and definite task completion window, and its time sequence-based power curve presents a block shape or high-low power consumption distribution. The interruptible load can be stopped and restarted at any time without additional cost, so it can be off-grid from the power grid under certain conditions or conditions. The historical power sequence of the interruptible load suddenly drops to zero at certain time periods. The power of the rigid load varies little and is continuously turned on, so the embodiment divides the load end to obtain the rigid load and the non-rigid load by analyzing the power segmentation of each load end on the historical power sequence.

[0028] Preferably, the off-grid condition of each load end is analyzed by using the historical power sequence, and a plurality of non-rigid loads are divided from the load end, and the specific steps of obtaining each working power period and the pre-interval length of each non-rigid load are as follows: The average value of the power of all time points in the historical power sequence of the i-th load end is recorded as the historical power average value of the i-th load end; A preset off-grid coefficient is taken as an example in the embodiment, and the product of the historical power average value of the i-th load end and the off-grid coefficient is recorded as the low-power threshold of the i-th load end; All time points in the historical power sequence of the i-th load end, at which the power is less than the low-power threshold, are recorded as the low-power time points of the i-th load end; According to the proportion of the number of low-power time points of each load end in the historical power sequence of the load end, all load ends are divided into non-rigid loads and rigid loads, and each working power period and the pre-interval length of each non-rigid load are obtained.

[0029] Specifically, according to the proportion of the number of low-power time points of each load end in the historical power sequence of the load end, all load ends are divided into non-rigid loads and rigid loads, and the specific steps of obtaining each working power period and the pre-interval length of each non-rigid load are as follows: The non-rigid load performance degree of the ith load end The calculation method is as follows: Wherein, is the time length of the historical power sequence, is the number of low-power time points in the historical power sequence of the ith load end; A preset rigid load screening threshold is taken as an example in this embodiment, and 0.5 is taken as an example for description; If the non-rigid load performance degree of the ith load end is greater than or equal to the rigid load screening threshold, the ith load end is recorded as a non-rigid load, each interval composed of non-low-power time points in the historical power sequence of the non-rigid load is recorded as a work power time period of the non-rigid load, and the time length between the last time point of the previous work power time period and the first time point of each work power time period in the historical power sequence of the non-rigid load is recorded as the pre-interval length of each work power time period of the non-rigid load. If the transferable load attribution degree of the ith load end is less than the rigid load screening threshold, the ith load end is recorded as a rigid load.

[0030] It should be noted that, since the non-rigid load includes interruptible load and transferable load, when the electric energy of the energy storage end in the power grid is insufficient to support all non-rigid loads after supplying the rigid load, the interruptible load needs to be disconnected from the power grid first, and the transferable load needs to be supplied according to the remaining electric quantity. In this process, the interruptible load is allowed to be directly disconnected from the power grid without considering the disconnection time and the power supply compensation in the next work power segmentation, while the transferable load needs to consider the interval of the disconnection time, so as to dynamically transfer the transferable load into the power grid, and supply and compensate the transferable load within the longest interval of each transferable load. Since the longest disconnection interval of different transferable loads is different, the longest disconnection interval of the transferable load is ignored in the adjustable resource regulation aiming at the lowest total operation cost, which causes safety risk of the transferable load.

[0031] Based on the above, since the interruptible load can be disconnected from the power grid over time, the distribution of the pre-interval length of the work power segmentation of the interruptible load has no regularity and predictability, while the transferable load has a maximum interval due to the existence of a definite task completion time window and a time limit requirement for task completion, so the time interval of the transferable load from disconnection from the power grid to reconnection into the power grid has regularity and aggregation. Therefore, the disconnection span aggregation degree of each non-rigid load is obtained by analyzing the length aggregation of the pre-interval length of all work power time periods of each non-rigid load.

[0032] Preferably, according to the consistency of all the pre-interval durations, the specific way to obtain the off-grid span concentration of each non-rigid load is: For the jth non-rigid load, the pre-interval durations of all the work power periods of the jth non-rigid load are counted, the kurtosis of all the pre-interval durations is obtained, denoted as ; The range of the pre-interval durations of all the work power periods of the jth non-rigid load is denoted as ; The calculation way of the off-grid span concentration of the jth non-rigid load is: It should be noted that the kurtosis and the range are existing known technologies, and the embodiment will not be described again; in the embodiment, the kurtosis indicates the concentration degree of the pre-interval durations, and the greater the value is, the more the pre-interval durations tend to the mean value of all the pre-interval durations of the jth non-rigid load, that is, the time interval of the ith non-rigid load from the grid each time is approximately, and the jth non-rigid load tends to be a transferable load rather than an interruptible load; the range indicates the maximum minimum difference of the pre-interval durations, which is used to limit the influence of discrete points when calculating the kurtosis, and the smaller the range is, the smaller the off-grid interval is, and thus the off-grid span concentration is greater, wherein, if all the pre-interval durations of the jth non-rigid load are equal, the range is 0, and the off-grid span concentration of the embodiment is calculated as 1.

[0033] It should be noted that the transferable load is mainly composed of air conditioners, heat pumps, electric heating loads, and AC loads with variable frequency drives, and thus the power of the transferable load has obvious periodic fluctuations. Compared with the interruptible load, the transferable load is made at the sacrifice of certain total operation cost, and thus when the transferable load is reconnected to the grid, the transferable load needs to be compensated according to the off-grid time length. For example, the transferable load is off-grid during the high-price period of the grid, the transferable load is reconnected to the grid during the low-price period of the grid, and the transferable load needs to be compensated in cost during the process, that is, the transferable load is supplied with power for a long time during the low-price period, and thus for each work power period of the transferable load, the longer the pre-interval duration of the work power period is, the longer the work power period is or the more power is consumed in the work power period, that is, the total power consumed by the transferable load in the work power period is in a positive proportional relationship with the time length of the pre-interval duration, so as to reflect the compensation mechanism of the grid to the transferable load when the grid performs power resource scheduling.

[0034] Based on the above, the embodiment obtains the transferable load attribution degree of each non-rigid load by analyzing the covariant relationship between the time length of each non-rigid load's work power period and the preceding interval length, and the covariant relationship between the total electric energy consumed by each work power period and the preceding interval length, and combining the off-grid span aggregation degree of each non-rigid load.

[0035] Preferably, the specific steps for obtaining the transferable load attribution degree of each non-rigid load by using the covariant relationship between each preceding interval length and the total amount of electric energy consumed for compensating power supply in the corresponding work power period are as follows: A compensation analysis space of the jth non-rigid load is constructed, and the compensation analysis space is a three-dimensional space, wherein the X-axis is the time length of each work power period of the jth non-rigid load, the Y-axis is the preceding interval length of each work power period of the jth non-rigid load, and the Z-axis is the output total electric energy of each work power period of the ith non-rigid load, which is expressed as the integral of time over power in each work power period of the jth non-rigid load; The time length, preceding interval length and output total electric energy of all work power periods of the jth non-rigid load are mapped into the compensation analysis space to obtain a plurality of sample points in the compensation analysis space of the jth non-rigid load; The least square method is used to linearly fit all sample points in the compensation analysis space to obtain a compensation straight line of the ith non-rigid load; The fitting correlation degree of each sample point in the compensation analysis space of the jth non-rigid load and the compensation straight line is analyzed to obtain a compensation performance degree of the jth non-rigid load; The transferable load attribution degree of the jth non-rigid load is obtained according to the compensation performance degree of the jth non-rigid load and the off-grid span aggregation degree of the jth non-rigid load, wherein the transferable load attribution degree is in a positive proportional relationship with the compensation performance degree and the off-grid span aggregation degree.

[0036] Specifically, the fitting correlation degree of each sample point in the compensation analysis space of the jth non-rigid load and the compensation straight line is analyzed to obtain the calculation method of the compensation performance degree of the jth non-rigid load. The compensation performance degree of the jth non-rigid load The specific method is as follows: Wherein, is the number of work power periods of the jth non-rigid load, is the Euclidean distance between the kth work power period corresponding to the sample point in the compensation analysis space of the jth non-rigid load and the compensation straight line of the jth non-rigid load; is an exponential function with a natural constant as a base, used for inversely proportional normalization of the accumulated sum of the Euclidean distances.

[0037] It should be noted that in the compensation analysis space, the closer the distance of the sample point to the compensation straight line, the closer it is to the fitting trend of all sample points, and the linear relationship represents the synergistic relationship, so the smaller the accumulated sum of the Euclidean distances, the stronger the covariant relationship, that is, the greater the value of the compensation performance degree.

[0038] Further, according to the compensation performance degree of the jth non-rigid load and the off-grid span concentration of the jth non-rigid load, the specific calculation method of the transferable load attribution degree of the ith non-rigid load is: The transferable load attribution degree of the jth non-rigid load The calculation method is: Wherein, is the off-grid span concentration of the jth non-rigid load.

[0039] Step S003, analyze the complexity of the power grid topology network to obtain the power supply cost of each non-rigid load for all energy storage ends; use the transferable load attribution degree and the power supply cost to obtain the power supply priority of each non-rigid load; use the power supply priority and the state of charge of all energy storage ends at the current time for model prediction to obtain the decision variable of each non-rigid load at each time period in the prediction time domain.

[0040] It should be noted that since the traditional adjustable resource scheduling is to obtain the decision variable with the purpose of minimizing the total operation cost, after predicting the future output of the energy storage end, the total power generation is equal to the total power consumption according to the total electricity demand of the rigid load and the non-rigid load, so as to determine the on-off state of each non-rigid load in the prediction time domain. However, the state of different energy storage ends and demand ends in the future power grid changes at any time, and only the supply-demand relationship obtains the decision variable without considering the power supply cost of each non-rigid load and the maximum interval duration of the transferable load in the non-rigid load. Therefore, the embodiment analyzes the power supply cost of each non-rigid load for all energy storage ends, and obtains the power supply priority of each non-rigid load by combining the transferable load attribution degree of each non-rigid load belonging to the transferable load.

[0041] It needs to be further explained that the power supply from the energy storage end to each demand end in the power grid is through line transfer of electric energy, the line itself has internal resistance, and when the electric energy is transferred from the end to the end of the power grid, it will be affected by the branches and intermediate demand ends. The more equipment and wiring the line passes through, the more reactive power loss of the electric energy, that is, the higher the power supply cost. Therefore, the embodiment obtains the power supply cost of each non-rigid load by analyzing the line complexity of all energy storage ends to the same non-rigid load in the power grid topology network.

[0042] Preferably, the complexity of the power grid topology network is analyzed to obtain the power supply cost of each non-rigid load by all energy storage ends; and the specific steps of obtaining the power supply priority of each non-rigid load by using the transferable load attribution and the power supply cost are as follows: In the power grid topology network, for the jth non-rigid load and the mth energy storage end, the shortest path from the mth energy storage end to the jth non-rigid load is obtained. The way of obtaining the shortest path is a known technology, and the shortest path is a node sequence from each energy storage end to the jth non-rigid load. The cost contribution coefficient of the mth energy storage end to the jth non-rigid load The calculation method is as follows: Wherein, The number of nodes on the shortest path from the mth energy storage end to the jth non-rigid load; The state of charge of the mth energy storage end at the current moment; The cumulative sum of degrees between all adjacent nodes on the shortest path from the mth energy storage end to the jth non-rigid load; It is a known technology, and the maximum-minimum normalization algorithm is taken as an example to normalize the cost contribution coefficient.

[0043] It needs to be explained that the number of nodes on the shortest path Indicates the number of branches and wiring on the path from the mth energy storage end to the jth non-rigid load. The larger the value, the higher the loss, and the larger the cost contribution coefficient. The value of the state of charge The larger the value, the higher the energy stored in the mth energy storage end at the current moment, and the higher the output power that can be provided to other demand ends in the prediction time domain. Therefore, it is the main power supply energy storage end even with certain loss due to its high state of charge. The cumulative sum of degrees indicates the length of the line passed through by the shortest path from the mth energy storage end to the jth non-rigid load. In the ideal case of the same line resistance, the longer the length of the line passed through by the electric energy, the higher the cost of power supply. The cost contribution coefficient of all energy storage terminals to the j-th non-rigid load. The average value is denoted as the power supply cost of all energy storage terminals to each non-rigid load. ; Using the transferable load attribution degree and the power supply cost, the power supply priority of each non-rigid load is obtained. The power supply priority is directly proportional to the transferable load attribution degree and indirectly proportional to the power supply cost.

[0044] As an example, in this embodiment, the ratio of the transferable load attribution degree of the j-th non-rigid load to the power supply cost of the j-th non-rigid load is denoted as the power supply priority of the j-th non-rigid load.

[0045] Furthermore, the specific steps for using the power supply priority and the state of charge of all energy storage terminals at the current moment to perform model prediction and obtain the decision variables for each non-rigid load in each time period within the prediction time domain are as follows: Defining the prediction window: In this embodiment, the prediction time domain is taken as 24 hours. The prediction time domain is divided into 15-minute segments to obtain 96 time periods. Determine the current state: Obtain the current state of charge of all energy storage terminals; obtain the current power of each rigid load; obtain the current power of each non-rigid load; obtain the electricity price for each time period in the prediction time domain; The optimization objective is set as follows: minimizing the total operating cost (electricity purchase cost) of the power grid. Construct a mixed-integer linear programming model and define binary decision variables. If the decision variable of each demand side is 1 in each time period within the prediction time domain, it means that the demand side is connected to the grid in that time period. If the decision variable of the demand side is 0 in that time period, it means that the demand side is disconnected from the grid in that time period. Construct the objective function : in, To predict the electricity price in time period a within the time domain, For the first The electricity purchase capacity of each demand side in the a-th time period within the forecast time domain; For the first The power supply priority of each demand side, wherein the demand side includes rigid loads and non-rigid loads, and the power supply priority of rigid loads is denoted as 1 in this embodiment; The mixed-integer linear programming model uses the branch and bound method to apply the objective function. Solving for the objective function to minimize the value while satisfying all constraints yields the decision variables for each energy storage unit in each time period of the prediction domain, i.e., the decision variables for each non-rigid load.

[0046] It should be noted that the solution to the optimal solution and the decision variables of the mixed integer linear programming model are existing and well-known techniques. The objective function described in this embodiment is only used to achieve the optimization objective of minimizing the total operating cost of the power grid. Other embodiments can achieve other optimization objectives by changing the current state and reconstructing the objective function. This embodiment does not impose any specific limitations.

[0047] Step S004: In each time period within the prediction time domain, the difference between the off-grid duration of each non-rigid load and its preceding interval duration is analyzed using decision variables. Combined with the real-time state of charge of the energy storage terminal, the decision update variables for each non-rigid load in each time period within the prediction time domain are obtained. The power grid uses the decision update variables to determine the grid connection and off-grid of each non-rigid load in each time period within the prediction time domain.

[0048] It should be noted that as the time period within the prediction domain increases, if the same energy storage terminal supplies power for an extended period, the state of charge of the energy storage terminal will decrease, which will be insufficient to support continuous power supply to non-rigid loads. At the same time, since transferable loads among non-rigid loads usually have the longest off-grid interval, if the off-grid time of transferable loads among non-rigid loads is too long, it will lead to a decrease in the stability of their operation. Therefore, this embodiment analyzes the interruption time of each transferable load and adjusts the power supply priority of each non-rigid load in each time period within the prediction domain by combining the transferable load affiliation degree, thereby obtaining the adjustment priority of each non-rigid load in each time period within the prediction domain.

[0049] Preferably, in each time period within the prediction time domain, the difference between the off-grid duration of each non-rigid load and its preceding interval duration is analyzed using decision variables. The power supply priority of each non-rigid load in each time period within the prediction time domain is adjusted, and the specific steps for obtaining the adjustment priority of each non-rigid load in each time period within the prediction time domain are as follows: Let the average of the preceding interval durations for all power output periods of the j-th non-rigid load be denoted as . ; Let any time period other than the first time period in the prediction time domain be denoted as the target time period q; The adjustment priority of the j-th non-rigid load in the target time period q within the forecast time domain is calculated as follows: in, Let be the number of time periods in the forecast time domain before the q-th time period where the decision variable is 1 for the j-th non-rigid load. Let be the transferable load attribution degree of the j-th non-rigid load; The adjustment priority of the j-th non-rigid load in the time period preceding the target time period within the forecast time domain. The preset duration of the time period is 15 minutes in this embodiment.

[0050] Furthermore, by using the method of obtaining the decision variables of each non-rigid load in each time period within the prediction time domain, the real-time state of charge of all energy storage terminals in each time period within the prediction time domain and the adjustment priority are used as inputs to obtain the decision update variables of each non-rigid load in each time period within the prediction time domain. The power grid uses the decision update variables to determine the grid connection and disconnection of each non-rigid load in each time period within the prediction time domain.

[0051] It should be noted that the embodiments used in this example The model only represents negative correlations and constraints. The model output results are in... Within the interval, This is the input to this model; in specific implementations, it can be replaced with other models that have the same purpose. This embodiment is merely an example. The description will be based on a model, without making any specific limitations.

[0052] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A demand-side adjustable resource response optimization method based on model predictive control theory, characterized in that, The method includes the following steps: Obtain the historical power sequence of each load and the state of charge of each energy storage unit at the current moment, and construct the power grid topology network of the load and energy storage units based on their layout in the power grid. By analyzing the off-grid status of each load end using historical power series analysis, several non-rigid loads are obtained by dividing the load ends, and the power output period and its preceding interval length for each non-rigid load are obtained; based on the consistency of all the preceding interval lengths, the off-grid span clustering degree of each non-rigid load is obtained; the transferable load affiliation degree of each non-rigid load is obtained by using the covariance relationship between each preceding interval length and the total electrical energy consumed by the compensated power supply during the corresponding power output period; The complexity of the power grid topology is analyzed to obtain the power supply cost of each energy storage terminal to each non-rigid load; the power supply priority of each non-rigid load is obtained by using the transferable load affiliation degree and the power supply cost; the power supply priority and the state of charge of all energy storage terminals at the current time are used to perform model prediction to obtain the decision variables of each non-rigid load in each time period in the prediction time domain. In each time period within the prediction time domain, the difference between the off-grid duration of each non-rigid load and its preceding interval duration is analyzed using decision variables. Combined with the real-time state of charge of the energy storage terminal, the decision update variables for each non-rigid load in each time period within the prediction time domain are obtained. The power grid uses the decision update variables to determine the grid connection and off-grid of each non-rigid load in each time period within the prediction time domain.

2. The demand-side adjustable resource response optimization method based on model predictive control theory according to claim 1, characterized in that, The process of analyzing the off-grid status of each load end using historical power sequences, dividing the load ends into several non-rigid loads, and obtaining the power consumption period and its preceding interval for each non-rigid load includes: The average power at all times in the historical power sequence of the i-th load end is denoted as the historical power average of the i-th load end. A preset off-grid coefficient is used, and the product of the historical average power of the i-th load end and the off-grid coefficient is recorded as the low power threshold of the i-th load end. In the historical power sequence of the i-th load end, all times when the power is less than the low power threshold are recorded as the low power times of the i-th load end. Based on the proportion of low-power moments at each load end in the historical power sequence of that load end, all load ends are divided into non-rigid loads and rigid loads, and the power period of each non-rigid load and its preceding interval duration are obtained.

3. The demand-side adjustable resource response optimization method based on model predictive control theory according to claim 2, characterized in that, The process of dividing all load terminals into non-rigid loads and rigid loads based on the proportion of low-power moments at each load terminal in the historical power sequence of that load terminal, and obtaining the power period of each non-rigid load and its preceding interval duration includes: Non-rigid load performance at the i-th load end The calculation method is as follows: in, The time length of the historical power sequence. This represents the number of low-power moments in the historical power sequence of the i-th load end; A preset rigid load screening threshold is set. If the non-rigid load performance of the i-th load end is greater than or equal to the rigid load screening threshold, the i-th load end is recorded as a non-rigid load. Each interval formed by non-low power moments in the historical power sequence of the non-rigid load is recorded as a power period of the non-rigid load. The time length between the last moment of the previous power period and the first moment of each power period in the historical power sequence of the non-rigid load is recorded as the pre-interval duration of each power period of the non-rigid load.

4. The demand-side adjustable resource response optimization method based on model predictive control theory according to claim 1, characterized in that, The step of obtaining the off-grid span clustering degree for each non-rigid load based on the consistency of all the aforementioned pre-interval durations includes: For the j-th non-rigid load, the pre-interval duration of all power output periods of the j-th non-rigid load is calculated, and the kurtosis of all such pre-interval durations is obtained, denoted as . ; The range of the preceding interval duration for all power output periods of the j-th non-rigid load is denoted as . ; The calculation method for the off-grid span aggregation degree of the j-th non-rigid load is as follows: .

5. The demand-side adjustable resource response optimization method based on model predictive control theory according to claim 1, characterized in that, The method of obtaining the transferable load attribution degree of each non-rigid load by utilizing the covariance relationship between the duration of each pre-interval and the total electrical energy consumed by the compensated power supply during the corresponding power period includes: Construct a compensation analysis space for the j-th non-rigid load, where the X-axis represents the time length of each power-period of the j-th non-rigid load, the Y-axis represents the pre-interval time length of each power-period of the j-th non-rigid load, and the Z-axis represents the total output energy of each power-period of the i-th non-rigid load. The total energy is represented by the integral of time with respect to power within each power-period of the j-th non-rigid load. The time length, pre-interval length, and total output power of all power periods of the j-th non-rigid load are mapped to the compensation analysis space to obtain several sample points in the compensation analysis space of the j-th non-rigid load. The least squares method is used to perform linear fitting on all sample points in the compensation analysis space to obtain the compensation line for the i-th non-rigid load. Analyze the degree of fitting correlation between each sample point in the compensation analysis space of the j-th non-rigid load and the compensation line to obtain the degree of compensation performance of the j-th non-rigid load. The transferable load attribution degree of the j-th non-rigid load is obtained based on the compensation performance degree of the j-th non-rigid load and the off-grid span aggregation degree of the j-th non-rigid load, wherein the transferable load attribution degree is positively proportional to the compensation performance degree and the off-grid span aggregation degree.

6. The demand-side adjustable resource response optimization method based on model predictive control theory according to claim 5, characterized in that, The analysis of the fitting correlation between each sample point in the compensation analysis space of the j-th non-rigid load and the compensation line, and the acquisition of the compensation performance of the j-th non-rigid load, includes: The degree of fitting correlation between each sample point in the compensation analysis space of the j-th non-rigid load and the compensation line is analyzed. The calculation method for obtaining the compensation performance of the j-th non-rigid load is as follows: The degree of compensation performance of the j-th non-rigid load The specific method is as follows: in, Let be the number of time intervals during which the j-th non-rigid load performs its power output. In the compensation analysis space of the j-th non-rigid load, the Euclidean distance between the sample point corresponding to the k-th power generation period and the compensation line of the j-th non-rigid load is given. It is an exponential function with the natural constant as its base, used to inversely normalize the sum of the Euclidean distances.

7. The demand-side adjustable resource response optimization method based on model predictive control theory according to claim 1, characterized in that, The analysis of the complexity of the power grid topology network is used to obtain the power supply cost of all energy storage terminals to each non-rigid load. The power supply priority of each non-rigid load is obtained using the transferable load affiliation and the power supply cost, including: Find the shortest path from the m-th energy storage terminal to the j-th non-rigid load in the power grid topology; The cost contribution coefficient of each energy storage terminal to the j-th non-rigid load is obtained based on the shortest path. The average of the cost contribution coefficients of all energy storage terminals to the j-th non-rigid load is denoted as the power supply cost of all energy storage terminals to each non-rigid load. ; Using the transferable load attribution degree and the power supply cost, the power supply priority of each non-rigid load is obtained. The power supply priority is directly proportional to the transferable load attribution degree and indirectly proportional to the power supply cost.

8. The demand-side adjustable resource response optimization method based on model predictive control theory according to claim 7, characterized in that, The specific steps for obtaining the cost contribution coefficient include: The cost contribution coefficient of the m-th energy storage terminal to the j-th non-rigid load power supply The calculation method is as follows: in, This represents the number of nodes on the shortest path from the m-th energy storage terminal to the j-th non-rigid load. This represents the state of charge of the m-th energy storage device at the current moment. It is the sum of the degrees between all two adjacent nodes on the shortest path from the m-th energy storage terminal to the j-th non-rigid load; This is existing publicly known technology.

9. The demand-side adjustable resource response optimization method based on model predictive control theory according to claim 1, characterized in that, The step of analyzing the difference between the off-grid duration of each non-rigid load and its preceding interval duration using decision variables in each time period within the prediction time domain, and obtaining the decision update variables for each non-rigid load in each time period within the prediction time domain in conjunction with the real-time state of charge of the energy storage terminal, includes: In each time period within the prediction time domain, the difference between the off-grid duration of each non-rigid load and its preceding interval duration is analyzed using decision variables. The power supply priority of each non-rigid load in each time period within the prediction time domain is adjusted to obtain the adjustment priority of each non-rigid load in each time period within the prediction time domain. By using the method of obtaining the decision variables of each non-rigid load in each time period of the prediction time domain, the real-time state of charge of all energy storage terminals in each time period of the prediction time domain and the adjustment priority are used as inputs to obtain the decision update variables of each non-rigid load in each time period of the prediction time domain.

10. The demand-side adjustable resource response optimization method based on model predictive control theory according to claim 9, characterized in that, The specific steps for obtaining the adjustment priority include: Let the average of the preceding interval durations for all power output periods of the j-th non-rigid load be denoted as . ; Let any time period other than the first time period in the prediction time domain be denoted as the target time period q; The adjustment priority of the j-th non-rigid load in the target time period q within the forecast time domain is calculated as follows: in, Let be the number of time periods in the forecast time domain before the q-th time period where the decision variable is 1 for the j-th non-rigid load. Let be the transferable load attribution degree of the j-th non-rigid load; The adjustment priority of the j-th non-rigid load in the time period preceding the target time period within the forecast time domain. The preset duration for the time period.

Images